Abstract:

All designing a controlled experiment should begin with the design of the ideal experiment. If the ideal experiment cannot be executed because of financial, factual, moral, or legal obstacles, we should make a systematic effort to save the controlled character of the experiment by redesigning its objectionable features. As a rule, the redesigned experiment will be less powerful than the ideal one, and we must then decide whether we want to live with that loss or move to quasi-experimental designs. The review in this paper of redesign strategies that have saved controlled experiments in the past should help us to be prepared and inventive the next time our ideal experimental design runs into a roadblock.